A combinatorial approach to electronic healthcare records in pharmacoepidemiology

Abstract

There is a need to know when multi-country observational studies are useful and how they should be performed and interpreted, because there are no well-established methods to deal with the combination of data from several independent electronic healthcare registries. The overall aim of this thesis was to explore the value of using data from multiple countries when estimating the risk of fractures for patients with multiple sclerosis (MS) and the risk of fractures for patients exposed to thiazolidinediones (TZDs). Data were retrieved from three different electronic healthcare registries: the British General Practice Research Database (GPRD), the Dutch PHARMO Record Linkage System (RLS), and the Danish National Health Registers linked to the Danish MS Registry. Our studies of MS and fractures indicate that increased awareness of the risk of hip fracture is warranted in patients with MS. Attention should particularly be paid for MS patients aged 30–69 and for more disabled patients (i.e., who have a higher expanded disability status scale score). The use of antidepressants further increased fracture risks. The results for patients with MS exposed to glucocorticoids (GCs) were conflicting, although generally the link between GC use and fracture risk has been well established. A clinical risk score for fracture risk estimation may be a starting point for the communication of an individual’s fracture risk in patients with MS, such as the FRAX score or the score presented in one of the chapters of this thesis, which includes MS as a risk factor. A measurement of the bone mineral density may be the next step to determine if preventative treatment, such as bisphosphonates, strontium ranelate, raloxifene or denosumab, is indicated. Other possible interventions are the intake of calcium and vitamin D tablets, and the prevention of falls. Treatment with TZDs in diabetic patients has been associated with a relatively small increased fracture risk for women. If a patient has several other risk factors for fracture, such as a previous fracture, exposure to GCs, or a higher age, a fracture risk assessment may be considered. If this assessment shows an absolute fracture risk that is substantial, then discontinuation of TZD use may be an option. However, this involves a benefit-risk assessment between the best treatment option for diabetes and the risk of fractures. Multi-country studies add valuable insights to the field of observational research. Whether they are worth the effort depends on the specific exposure and outcome being investigated, the quality and completeness of data recording in the separate registries, the ability to gain access to individual patient data if needed and the willingness to invest time and money. When different registries are combined, a crucial step is to investigate the features of these separate registries and the characteristics of patients. Ideally, there should be as much consistency as possible between data from the different sources in terms of study design and definitions of exposure, outcome and risk factors before combining the data. Access to individual patient data is therefore a major strength for a multi-country study.